Food and Life, Pleasure and Worry, Among American College Students: Gender Differences and Regional Similarities
Page content transcription
If your browser does not render page correctly, please read the page content below
Journal of Personality and Social Psychology Copyright 2003 by the American Psychological Association, Inc. 2003, Vol. 85, No. 1, 132–141 0022-3514/03/$12.00 DOI: 10.1037/0022-35126.96.36.199 Food and Life, Pleasure and Worry, Among American College Students: Gender Differences and Regional Similarities Paul Rozin, Rebecca Bauer, and Dana Catanese University of Pennsylvania Questionnaires on food attitudes and behavior were completed by 2,200 American undergraduates from 6 regionally dispersed college campuses. Results indicate that a substantial minority of women and a much smaller minority of men have major concerns about eating and food with respect to both weight and health. Overall, 14% of women reported being embarrassed to buy a chocolate bar in the store. A 6-factor structure emerged: weight concern, diet and health orientation, beliefs about the diet– health link, food negativity/importance of food as a source of pleasure in life, eating disordered behaviors, and natural/ vegetarian food preferences. There were surprisingly few regional differences, virtually none if race and social class were taken into account. Gender was the strongest predictor of responses. There is little doubt that in contemporary American society, habits. However, the majority of individuals do not succeed, in the especially among women, food and eating lead to ambivalent long run, in reducing weight substantially (Wadden & Foster, feelings. The pleasure and necessity of eating are opposed by 2000). Therefore, concerns about diet with respect to health and concerns about appearance, the health effects of overweight, and weight may be nonproductive and produce worry, anxiety, and the risks of consuming a nonoptimal diet. Whatever one’s views failure, with consequences in terms of mood and self-esteem about the importance of a healthy diet, it is clear that many (Polivy & Herman, 2002). A number of authors have noted with Americans spend a lot of time worrying about calories and fat and alarm that obesity is increasing in the United States at the same that this concern, in itself, detracts from the quality of life (Rozin, time that more and more attention seems to be devoted to weight- Fischler, Imada, Sarubin, & Wrzesniewski, 1999). For a small loss programs (Wadden, Brownell, & Foster, 2002). minority of the population, especially women, these concerns This article addresses the food and eating concerns of Ameri- come to dominate life in the form of anorexia nervosa or in the cans in a broader context than is typically done in the literature. It various forms of bulimia. The focus of most past research in this is modeled after a prior study of food and life in four cultures area has been on the eating disorders rather than what Rodin, (United States, France, Flemish Belgium, and Japan; Rozin et al., Silberstein, and Striegel-Moore (1985) called the “normative dis- 1999). This broader context has been expressed outside of psy- content” with one’s body that is common among American women chology, as in the literatures in sociology (e.g., Fischler, 1990) and (p. 267). The focus of most past research has also been on calories, in more humanistic and historical approaches (e.g., Kass, 1994; that is, amount consumed, and accompanying weight problems, Stearns, 1997; for a general discussion of American food ambiv- leaving aside important worries about quality and composition of alence, see Rozin, 1998). diet. The common American avoidance of high-fat foods seems to We are particularly motivated to explore what we may term an be more than just an avoidance of calories; fat seems to have ambivalence about food and eating in many Americans because assumed, even at low levels, the role of a toxin. Thus, almost a this attitude seems to reduce the quality of life and to be unpro- third of a sample of Americans seek a totally fat-free (and hence, ductive. The French are at least as healthy as Americans and have actually fatal) diet (Rozin, Ashmore, & Markwith, 1996). a much lower incidence of heart disease (Renaud & de Lorgeril, The normative discontent about both weight and food choice is 1989), yet consume a diet somewhat higher in fat (Drewnowski et particularly problematic because in both cases, it seems to lay- al., 1996) and have a much more relaxed attitude to eating (Rozin people that there should be little difficulty in changing eating et al., 1999). The French seem to focus on the experience of eating, whereas Americans focus more on the consequences of eating (Rozin et al., 1999). The French probably eat somewhat less than Paul Rozin, Rebecca Bauer, and Dana Catanese, Department of Psy- Americans but get more food experience because they pay more chology, University of Pennsylvania. attention to the food they are eating (Rozin, Kabnick, Pete, This research was supported, in part, by funds from the Edmund J. and Fischler, & Shields, in press). Louise W. Kahn Chair for Faculty Excellence at the University of Penn- Unlike the European nations, the United States has minimal sylvania. We thank Leann Birch, Richard Davidson, Alan Fridlund, and language barriers and is influenced as a whole by substantially the Carol Nemeroff for helping to provide access to the students at the various same films, magazines, television, papers and legislation. These universities. The present article was written while Paul Rozin was a features may reduce regionalization. On the other hand, the United Visiting Scholar at the Russell Sage Foundation. Correspondence concerning this article should be addressed to Paul States spans a wide range of climates. Climate influences clothing, Rozin, Department of Psychology, University of Pennsylvania, 3815 with the clothing in warmer climes more likely to reveal body Walnut Street, Philadelphia, Pennsylvania 19104-6196. E-mail: rozin@ shape. Climate also influences degree of exercise and the avail- psych.upenn.edu ability of foods such as fruits. There are also substantial American 132
FOOD AND LIFE, PLEASURE AND WORRY 133 regional differences in cultural and ethnic backgrounds. People 42% middle class, and 48% upper middle or upper class. The predominant who derive from ancestors in Europe, Asia, Mexico or further race was Caucasian (72%), and the predominant religions were Catholic south in the Americas, and Africa are distributed very unequally (38%) and Protestant (37%). (See Table 1 for full sample distribution by around the country. Any discovered regional differences would race and religion.) The one-page, two-sided anonymous questionnaires received a very high have to be analyzed to determine whether the difference was based return rate because student questionnaires were filled out in introductory on region or ethnic–racial makeup. psychology classes or during required periods for filling out questionnaires The literature in psychology on eating disorders and food and associated with some introductory psychology courses. Feedback on results eating in general has focused on undergraduate students. Although was made available to the instructors of the classes that contributed results, this subsample of Americans is atypical in many ways, particularly when requested, within 1 month of data collection. with respect to food choice, it has the virtue of being a focal group The questionnaire consisted of a range of demographic items plus 69 for the appearance of eating disorders. Along with obesity, eating food-relevant items. These were primarily from the food-life questionnaire disorders are responsible for motivating almost all of the research in Rozin et al. (1999). Some of the questions were modified or improved on food in psychology. The research on undergraduates has been on the basis of problems encountered in prior analyses. A small number of surprisingly inattentive to variables other than gender that might items were open ended, but these are not included in this analysis. The questions used in the present analysis, except for self-reported weight and differentiate undergraduates. Location of a study (e.g., region of height, are true–false, frequency measures (e.g., never, rarely, sometimes, the United States); region of origin of the sample; and religion, often, almost always), or choices between two specific alternatives. A few religiosity, and social class of respondents are rarely mentioned, items represent what we consider “default” ways of thinking about food- and race– ethnic origin receives relatively little attention. In sum- related matters (Rozin et al., 1999). These items, like free associations, marizing the literature, especially in experimental studies, results probe the most salient way that one thinks about relevant issues. For tend to be considered as equivalent if derived from any American example, Rozin et al. (1999) reported that although Americans tend to (or Canadian) undergraduate samples. think of food in terms of its health or nutritional properties, the French tend In this study, we attempted to address and fill some of the gaps to think in terms of the experience of eating foods. One default item is in the literature that we have identified above: “Circle the word that you most readily associate with the word at the left”; 1. We explored, via questionnaire, a wide range of attitudes and the word is heavy cream, and the choices are whipped and unhealthy. On the actual questionnaire, items are organized by format of question (true– behaviors toward food, including issues such as default thinking about false, frequency, default, etc.). food, beliefs about diet and health, and concerns about amount and quality of food eaten. We built upon and modified another attempt to survey the role of food in life (Rozin et al., 1999). Results 2. For the first time, we addressed the justification for pooling Factor Analysis: The Structure of Food Attitudes and results across American regions: We compared undergraduate Behaviors introductory psychology samples from six universities spread across the United States. All variables were rescaled to vary from 0 to 1. Thus, an item on 3. We analyzed results in terms of gender, race, religion, and a 5-point scale was converted to scores of 0, .25, .50, .75, and 1.00. self-reported social class as well as region of the United States. Examination of the factor analysis (PCA, Varimax rotation, Our study on regional differences was prompted by a pilot study SYSTAT9 [SPSS, 1999]) of 58 items suggested that a six-factor by Bauer and Catanese (1996), which suggested much greater solution might be optimal, given the scree plot, and so a six-factor body concerns and dieting and diet– health concerns among un- solution was forced (accounting for 35% of the variance). Assign- dergraduate women at the University of Pennsylvania who were ment of items to factors was extremely clear; only four of the 58 from the West coast as opposed to those from the East coast. items did not load at least .30 on any factor, and only two items loaded at .40 or more on more than one factor. Table 2 organizes Method the items by factor (and the residual four items) and indicates the name of the factor and the items that loaded at least .40 on that Participants were 2,200 college students. We eliminated the data from 38 factor (with the value dropped to greater than .30 if this was the participants because these had more than 6 missing values across the highest loading that this item had on any factor). relevant 58 questionnaire items that were subject to analysis. For purposes of subsequent analysis, the missing values of the remaining participants Factor scores were created by combining items in accordance were assigned the modal value for the whole sample of the item in with their loading on each factor (Table 2). Only the items listed question. Of the remaining 2,162 participants, 59% were women. Partici- in Table 2 were used to compose each factor. Loadings less than pants were distributed across six geographically dispersed large universi- .30 were not included in the factor score, nor were loadings less ties: Universities of Pennsylvania, Texas at Houston, Wisconsin, and than .40 unless they were the highest loading on any factor. Scores California at Santa Barbara; Pennsylvania State University; and Arizona on individual items were reversed, as appropriate, so that high State University. Five of the six schools are state universities (the Univer- scores on each item “pointed” in the direction of concern about sity of Pennsylvania, in spite of its name, being the exception). State food. The factors extracted from this database are similar to, but universities have the advantages that they draw students disproportionately not identical to, those derived from a similar questionnaire used on from the region that they are in and that because of low tuition, there are college students and adults from four countries (Rozin et al., smaller barriers to entry by students from lower socioeconomic back- grounds. The six universities are widely dispersed in the United States: two 1999). from the Northeast and one each from the South, Midwest, Southwest, and We now consider the individual factors, beginning in each case California. with a brief interpretation of the meaning of the factor followed by The mean age of the participants was 19.3 years. Self-reported socio- a discussion of gender, school, race, religion, and social class economic class (5-point scale) yielded 10% lower or lower middle class, effects. For each of the factors, mean values for factor scores
134 ROZIN, BAUER, AND CATANESE Table 1 Participant Demographics Demographic ASU PSU UTEX UCSB UPENN UWIS Total No. 206 285 297 372 276 726 2,162 Age (M)a 19.9 20.4 19.3 18.8 19.2 18.9 19.3 Men n 94 54 112 129 117 371 877 % 46 19 38 35 42 51 41 Socioeconomic class Lower/lower middle n 19 29 56 35 25 47 211 % 9 11 19 10 9 7 10 Middle n 108 132 147 150 75 276 888 % 53 48 50 41 28 38 42 Upper middle/upper n 77 115 92 180 172 394 1,030 % 38 42 31 49 63 55 48 Religion Atheist/agnostic n 19 15 9 57 45 77 222 % 9 5 3 16 16 11 11 Buddhist n 5 2 28 10 6 6 57 % 3 1 10 3 2 1 3 Catholic n 76 133 112 116 71 295 803 % 39 48 38 33 26 42 38 Hindu n 0 0 12 0 16 3 31 % 0 0 4 0 6 1 1 Jewish n 6 18 4 29 69 52 176 % 3 6 1 8 25 7 8 Muslim n 2 2 18 3 10 9 44 % 1 1 6 1 4 1 2 Protestantb n 89 109 111 139 56 264 768 % 45 39 38 39 21 37 37 Race African American n 8 14 39 9 8 7 85 % 4 5 13 2 3 1 4 East/Southeast Asian, Pacific Islander n 14 4 71 55 43 30 217 % 7 1 24 15 16 4 10 Hispanic, Native American n 22 7 64 44 17 12 166 % 11 2 22 12 6 2 8 South Asian, Indian n 2 3 29 2 25 8 69 % 1 1 10 1 9 1 3 Caucasian n 152 251 82 231 176 646 1,538 % 75 89 28 64 64 90 72 Other n 4 3 9 19 5 15 55 % 2 1 3 5 2 2 3 Note. ASU ⫽ Arizona State University; PSU ⫽ Pennsylvania State University; UTEX ⫽ University of Texas; UCSB ⫽ University of California, Santa Barbara; UPENN ⫽ University of Pennsylvania; UWIS ⫽ University of Wisconsin. a Mean of male and female mean ages, for each campus. b Includes all sects of Christianity other than Catholicism.
FOOD AND LIFE, PLEASURE AND WORRY 135 by school and gender are shown in Table 3, and results of the hand, 48% of women and 74% of men said they consumed fast analyses of variance (ANOVAs) by gender, school–region, race food on at least a weekly basis. (Black, East/Southeast Asian, Hispanic, American Indian, South The major gender effects translate into a difference between a Asian, White, and other), and religion (atheist/agnostic, Buddhist, mean factor score of .22 for women and ⫺.33 for men, a very Catholic, Jewish, Hindu, Muslim, and Protestant) and correlations substantial .56 z-unit difference, but notably less than the differ- with self-reported social class are shown in Table 4. In discus- ence for Factor 1. The range of factor scores across schools and sion of results, we adopt the .01 level as a criterion for gender is substantial (0.92 z units), with University of Texas men significance. lowest (⫺.44) and University of Wisconsin women highest (.48). Factor 1: Weight concern. The items with highest loadings on This variable shows the broadest range of significant effects of this factor deal with concerns regarding weight and appearance variables other than gender. By far the most significant school and specific weight-concern-related behaviors (e.g., dieting; see effect, F(5, 2155) ⫽ 15.52, p ⬍ .001, occurs for this variable. It is Table 2). This factor accounted for more variance than any other. accounted for, principally, by the anomalous results from the Overall, there is a great deal of concern about weight in Amer- University of Texas women. Their factor score on this variable ican undergraduates: 20% of the sample of over 2,000 was cur- (⫺.41) is among the most negative of the male scores and different rently on a diet, a full 28% would willingly exchange eating for an in sign from all of the other female scores (Table 3). Post hoc tests inexpensive nutrient pill, 43% were concerned about their weight indicate that the female University of Texas Factor 2 score differs often or almost always, and 47% thought their thighs were too fat. significantly from that of the women at each other school by at Of all factors, weight concern shows the largest difference in factor scores across gender, with a mean female factor score of .43 least p ⬍ .01. A consequence of this anomaly is that in a two-way compared with a male score of ⫺.63. Women were more con- ANOVA, there is a substantial and significant Gender ⫻ School cerned with appearance and diet in each of the six schools: 58% of interaction term, F(1, 2149) ⫽ 6.44, p ⬍ .001. We devote a men responded “never” to “How often do you diet?” in compari- subsequent section of this article to the explanation of this most son with 33% of women; 72% of women reported that their “thighs anomalous result. were too fat” in comparison with 11% of men; 34% of women Significant effects of race, F(6, 2123) ⫽ 17.96, p ⬍ .001, and were happy with their present weight as opposed to 58% of men; religion, F(6, 2096) ⫽ 5.45, p ⬍ .001, both derive principally from 34% of women claimed a willingness to give up eating if it could the fact that Hindus and South Asians showed by far the lowest be replaced by an inexpensive nutrient pill as opposed to 21% of score on diet– health concern. The significant link to social class men. (r ⫽ .12) confirms the belief that diet– health concerns become In one-way ANOVAs, there was a highly significant effect of greater as social class increases. gender, F(1, 2160) ⫽ 797.75, p ⬍ .001; the only other significant A two-way ANCOVA for Gender ⫻ School, with race, religion, effect was a much more modest effect of religion, F(6, and class as covariates, produces a still massive, F(1, 2040) ⫽ 2096) ⫽ 4.77, p ⬍ .001 (Table 4). The principal outlier group on 127.05, p ⬍ .001, effect of gender and a still significant, F(1, religion was Hindus, who had a particularly low score. Factor 2040) ⫽ 6.04, p ⬍ .001, effect of school as well as a significant scores by gender and school varied considerably (range of z interaction of school and gender, F(1, 2040) ⫽ 5.05, p ⬍ .001. scores ⫽ 1.34), anchored by ⫺.73 for Pennsylvania State Univer- Hence, on this variable, there seem to be some substantial regional sity men and .61 for University of Wisconsin women. differences. A two-way school and gender analysis of covariance Factor 3: Belief in a diet– health linkage. These four items ask (ANCOVA), with race, social class, and religion as covariates, for beliefs about the diet– health link in general and for obesity, yields a still massive gender effect, F(1, 2140) ⫽ 580.24, p ⬍ .001, cancer, and heart disease. The four items form a distinctive factor, and much more modest significant effects of school, F(1, with the lowest factor loading of the four items at .65 (and the next 2140) ⫽ 3.08, p ⬍ .01, and a Gender ⫻ School interaction, F(1, highest item loading only .21). Overall, the American undergrad- 2140) ⫽ 3.17, p ⬍ .01. uates seemed convinced that diet has a substantial effect on health. Factor 2: Diet– health orientation. This factor focuses on con- For example, 61% of all participants believed diet has a “strong cerns about diet and health: modification of food choice to reduce effect” (the highest of four choices) on heart disease. fat and salt, and emphasis in choice and thought on nutrition and There is a substantial gender effect for this variable, F(1, health as opposed to taste or culinary properties. In the food choice 2160) ⫽ 53.63, p ⬍ .001. The gender effect occurs because literature, taste has emerged as the dominant factor in food choice, but 35% of the sample said nutrition was more important in their women tend to believe in a stronger diet– health link (mean factor food choice than taste, 24% of the sample ate foods from which fat score for women is .13 compared with ⫺.19 for men), but the had been removed at least a few times a week, and 60% associated major finding on this factor is that the great majority of our heavy cream more with unhealthy than whipped. More generally, participants believed in a strong diet– health link. The range of many in the sample viewed food more in nutritional terms than in scores on this variable, across school and gender (Table 3) is taste–pleasure terms. modest, covering .52 z units, from ⫺.24 for University of Texas The major effect for this variable is gender, F(1, 2160) ⫽ men to .28 for Pennsylvania State University women. 171.91, p ⬍ .001, with women higher on diet– health concern. For There are modest, but significant effects of race and school example, whereas 39% of women ate low-fat (fat-removed) foods (Table 4). A Gender ⫻ School two-factor ANCOVA, with race, at least daily, only 15% of men did. Whereas 68% of women religion, and class as covariates, reveals a remaining substantial associated heavy cream with unhealthy as opposed to the culinary effect of gender, F(1, 2040) ⫽ 32.42, p ⬍ .001, and no other choice, whipped, only 49% of men made this choice. On the other significant effects.
136 ROZIN, BAUER, AND CATANESE Table 2 Questionnaire Items, Conceptually Arranged, in Terms of the Scores Derived From the Six-Factor Analysis Scorea Factor Item name loadingb Item Female Male Factor 1: Weight concern CONCWGT .80R I am concerned about being overweight.c .65 .33 GUILT .76R I feel guilty when I overeat.c .62 .26 DIET .67R I am currently on a diet.c .41 .18 THIGH .66 My thighs are too fat. .72 .11 HOLDBACK .64R I consciously hold back at meal time, so as not to gain weight.c .38 .19 SELFWGT .64 Relative to your same sex peers, how do you perceive yourself?d .44 .36 BUTT .61 My butt is too fat. .44 .10 HAPWGT .60R Are you happy with your present weight? .65 .42 ARMS .58 My arms are too fat. .29 .03 CURRDIET .51 I am currently on a diet to lose weight. .27 .10 APPEAR .49 I am concerned about what I eat and how it will affect my appearance. .78 .51 BELLY .48 My belly is too fat. .65 .64 OBSWGT .48 I am obsessively preoccupied with my weight.e .29 .10 FACE .47 My face is too fat. .18 .08 CHOCCAKE .46R Circle the word that you most readily associate with the word at the left: .32 .12 chocolate cake guilt celebration PILL .41 If I could eliminate eating by satisfying nutritional needs, safely, cheaply, and without .34 .21 hunger, with a pill, I would. ICECREAM .37 Circle the word that you most readily associate with the word at the left: .32 .14 ice cream delicious fattening EXERFUWO .32 Circle the word that you most readily associate with the word at the left: .67 .49 exercise fun work Factor 2: Diet–health orientation LOWFAT .65R Frequency with which you eat low-fat food in which some of the fat has been removed .72 .50 or substituted for high fat food.f HEALTHEAT .64 I am a healthy eater. .64 .54 FASTFOOD .59R Frequency with which you eat fast food.f .64 .47 NUTTASTE .58R Which is more important in food choice: nutrition or taste? .40 .24 CONTCAL .50 Do you control your caloric intake? .38 .21 BREAD .49R Circle the word that you most readily associate with the word at the left: .75 .64 bread carbohydrate butter EXERCISE .48R Frequency with which you exercise?f .64 .68 LONGTERM .48R I rarely think about the long term effects of my diet on health. .69 .53 EDUCNUTR .43 I am educated about nutrition. .79 .75 HVCREAM .41 Circle the word that you most readily associate with the word at the left: .68 .49 heavy cream whipped unhealthy WALKTV .41R Do you prefer to walk or watch TV? .56 .36 FOODHEPLg .41R Circle the word that you most readily associate with the word at the left: .48 .37 food health pleasure LOWSALT .40R Frequency with which you eat reduced salt [food from which salt is removed] products?f .35 .27 MILK .38 Circle the word that you most readily associate with the word at the left: .76 .63 milk cookies calcium BROCCOLI .36 Circle the word that you most readily associate with the word at the left: .84 .75 broccoli butter vitamins SACCHAR .34R Use no, or very-low calorie sweeteners.f .30 .15 FREGG .33 Circle the word that you most readily associate with the word at the left: .57 .32 fried egg breakfast cholesterol VITAMIN .32R Take vitamins.f .43 .40 Factor 3: Diet–health link HEART .80 How much of an effect does diet have on heart disease?h .83 .80 GOODHEALTH .70 How much of an effect does diet have on good health?h .88 .84 OBESITY .68 How much of an effect does diet have on obesity?h .87 .80 CANCER .64 How much of an effect does diet have on cancer?h .60 .48 Factor 4: Food negativity ENJOYED .67R Enjoying food is one of the most important pleasures in my life. .42 .37 FONDMEM .57R I have fond memories of family food occasions. .35 .38 THINKPOS .53R I think about food in a positive anticipatory way. .31 .18
FOOD AND LIFE, PLEASURE AND WORRY 137 Table 2 (continued ) Scorea Factor Item name loadingb Item Female Male Factor 4: Food negativity (continued) MONEY .48 Money spent on food is well spent. .38 .20 FOODHEPL .45Rg Circle the world that you most readily associate with the word at the left: .48 .37 food health pleasure HOTEL .42 For the same price, would you rather vacation at an average hotel with the best food or .84 .73 a luxurious hotel with average food? Factor 5: Eating-disordered characteristics OBSEXER .58 I am obsessively preoccupied with my exercise habits. .19 .21 OBSWGT .45e I am obsessively preoccupied with my weight. .29 .10 PURGE .44R How often have you purged (vomited) intentionally after eating for the purpose of .06 .02 weight reduction?i BINGE .40 Frequency with which you binge?f .12 .16 EMBCHOC .38 I am embarrassed to buy a chocolate bar in the store. .14 .04 Factor 6: Natural/vegetarian DATESALAD .59 I would rather date someone who eats lots of fruits and veggies rather than lots of .17 .22 meats. NATTASTE .58 I think natural/organic foods are better tasting than commercially grown/processed foods. .35 .33 NATHEALTH .54 I think natural/organic foods are healthier than commercially grown/processed foods. .77 .74 VEGETAR .40R Are you a vegetarian?j .14 .04 CNDYNICE .37 People who eat lots of candy aren’t as nice as people who eat lots of fruits and .03 .06 vegetables. Items that load ⬍ .30 on any factor FRNDHLTH I am concerned with the health of a friend/family member who eats poorly. .74 .66 SUGAR R Circle the word that you most readily associate with the word at the left: .40 .48 sugar cavities pleasure HOLIDAY R Circle the word that you most readily associate with the word at the left: .41 .31 holiday dinner overeat happiness PASTA R Circle the word that you most readily associate with the word at the left: .32 .24 pasta bread sauce Note. Response alternatives to items are true–false or yes–no unless otherwise indicated or footnoted. All items loading at least .40 on any factor are listed under that factor, in order of decreasing loading, with two items listed twice. Items loading greater than .30 are also included if there was no loading greater than or equal to .40 on any factor, and they represent the highest loading of the item on any factor. The items listed under each factor are exactly those that contributed to the factor score, weighted by their loadings. 1.00 ⫽ the most food-concerned response; 0 ⫽ the most food-positive response. R ⫽ reverse scored. Items are adapted from Rozin et al. (1999). a Percentage of females and males giving the more food-concerned response (cut-off set at some intermediate value for items with more than two response alternatives). b Mean score on a 0 –1 scale, with 1 ⫽ the most health/worry-oriented alternative. Exercise scored as if it was food. c 1 ⫽ almost always, 2 ⫽ often, 3 ⫽ sometimes, 4 ⫽ rarely, 5 ⫽ never. d 1 ⫽ underweight to 5 ⫽ overweight. e Listed under both Factor 1 (WGTCONCERN) and Factor 5 (EATDIS). f 1 ⫽ every day, 2 ⫽ a few times a week, 3 ⫽ once a week, 4 ⫽ once or twice a month, 5 ⫽ rarely/never. g Listed under both Factor 2 (HEALTHCONCERN) and Factor 4 (FOODNEGATIVE). h 1 ⫽ no effect, 2 ⫽ little effect, 3 ⫽ some effect, 4 ⫽ strong effect. i Never, Once/Twice, Occasionally, Often, Almost daily. j Vegan, No meat or fish, Partial, Not at all. Factor 4: Food negativity (food importance; positive–negative). showed a 58% female and 63% male endorsement, whereas 38% This distinct factor is principally composed of seven items and of women as opposed to 20% of men disagreed that “money spent has to do with the degree to which feelings about food are on food is well spent.” The highest (most food-negative) factor negative (or inversely, the positivity and importance of food in score was from Arizona State University women (.33), and the life). Overall, 60% of respondents claimed that enjoying food lowest was from University of Pennsylvania men (⫺.32), yielding was one of the greatest pleasures in their lives, but only 21% a range of .65 z units. chose a hotel with gourmet food and average accommodations There are modest effects of other demographic variables, the over a hotel, at the same price, with average food and luxurious most substantial being race, F(6, 2123) ⫽ 3.98, p ⬍ .001; the most accommodations, and 28% would prefer a nutrient pill to eat- striking racial difference is the highest positive reaction to food ing. A total of 44% associated food more with health than with among East and Southeast Asians. The gender effect remains pleasure. substantial, F(1, 2040) ⫽ 42.89, p ⬍ .001, in the two-way Gen- This factor shows the expected major effect of gender, by far the der ⫻ School ANCOVA, with the other demographic variables as most significant variable, F(1, 2160) ⫽ 51.24, p ⬍ .001. The covariates. There is no significant effect of school or the interac- highest loaded item, ENJOYED (cited in the previous paragraph) tion in this ANCOVA.
138 ROZIN, BAUER, AND CATANESE Table 3 Factor Summary of Data on Selected Variables by School and Gender Item ASU PSU UTEX UCSB UPENN UWIS n M 94 54 112 129 117 371 F 112 231 185 243 159 355 Factors 1. Weight concern M ⫺64 ⫺73 ⫺54 ⫺71 ⫺63 ⫺61 F 40 36 19 40 46 61 2. Health orientation M ⫺39 ⫺32 ⫺44 ⫺40 ⫺20 ⫺30 F 12 30 ⫺41a 19 41 48 3. Diet–health beliefs M ⫺24 4 ⫺22 ⫺20 ⫺23 ⫺18 F 2 28 3 3 11 19 4. Food negativity M ⫺16 ⫺18 ⫺10 ⫺24 ⫺32 ⫺15 F 33 1 8 8 10 18 5. Eating disorder M ⫺16 ⫺22 6a ⫺10 ⫺28 ⫺17 F 24 4 0 9 17 12 6. Natural/vegetarian M ⫺12 ⫺3 27a 12 0 ⫺16 F ⫺2 ⫺19 9 8 16 4 Mean of Factors 1–5 M ⫺28 ⫺24 ⫺16 ⫺26 ⫺28 ⫺26 F 18 13 0 14 24 27 BMI M 24.0 24.3 23.5 23.4 23.1 23.7 F 22.0 22.2 21.8 21.6 21.3 22.2 Note. Scores on each factor are multiplied by 100. Positive scores indicate greater concern/worry about food and eating. ASU ⫽ Arizona State University; PSU ⫽ Pennsylvania State University; UTEX ⫽ University of Texas; UCSB ⫽ University of California, Santa Barbara; UPENN ⫽ University of Pennsylvania; UWIS ⫽ University of Wisconsin; M ⫽ male; F ⫽ female; BMI ⫽ body mass index (weight kg/height m2). a Most anomalous UTEX findings. Factor 5: Eating disorders. This factor groups five items that The principal predictor on this factor is gender, F(1, ask specifically about disordered eating behaviors such as the 2160) ⫽ 31.88, p ⬍ .001, with women showing higher (more frequency of purging and binging and obsessions about eating and disordered) scores. Most strikingly, the incidence of embarrass- exercising. The sample shows substantial presence of strong con- ment about buying a chocolate bar in the store was higher in cern about food: Overall, 20% of the sample claimed to obsess women (14% for women, 4% for men). Incidences of binging, about exercise and 21% to obsess about eating; 89% had never purging, and dieting were higher among women; 15% of women purged, and 67% had never binged. admitted to having purged on at least one occasion, which con- Table 4 One-Way Analyses of Variance or Correlations Between Factors or Body Mass Index (BMI) and Demographic Variables Item Gender School Religion Race Class (Pearson r) Factor 1. Weight concern 797.75** 2.15 4.77** 1.67 .02 2. Health orientation 171.91** 15.52** 5.45** 17.96** .12** 3. Diet–health link 53.63** 4.00** 2.65 4.19** .02 4. Food negativity 51.24** 0.84 2.97* 3.98** ⫺.01 5. Eating disorder 31.88** 0.34 2.11 0.94 .03 6. Natural/vegetarian 1.47 4.71** 7.64** 2.76 .01 Factor average 328.43** 1.78 2.89* 5.76** .05 BMI 131.14** 3.73* 4.03** 12.35** ⫺.06 Note. n: gender, school ⫽ 2,162; BMI ⫽ 2,120; race ⫽ 2,130; religion ⫽ 2,103; class ⫽ 2,087. * p ⬍ .01. ** p ⬍ .001.
FOOD AND LIFE, PLEASURE AND WORRY 139 trasts with the 4% of men who had purged on at least one occasion. least some of this difference could result from these variables as Whereas 28% of females classified themselves as obsessed with opposed to region. Using the University of Wisconsin data as their weight, only 11% of men did. The highest factor score on representative of the rest of the sample, we recalculated the factor eating disorders is Arizona State University women (.24), and the scores using only respondents who were White and at least middle lowest is University of Pennsylvania men (⫺.28). Gender is the class. This substantial narrowing of the sample, along the two most only significant effect revealed on this factor (Table 4). promising demographic characteristics, reduced the University of Factor 6: Natural/vegetarian. This factor contains a number Texas anomaly by about 50%; that is, on health concerns, Univer- of items reflecting preference for natural foods, moral and other sity of Texas women move from negative to about neutral, and on implications of natural food preferences, and vegetarian status. It is natural/vegetarian, University of Texas men move close to neutral, the only factor that does not show a significant gender effect, even but still on the “female,” pronatural side. Examination of the though it includes a vegetarian item that confirms the higher scores of the University of Texas sample on the individual items incidence of vegetarianism in women (including rejection of red that make up these factors reveals that the anomalies appear rather meat: 34% of women vs. 10% of men). Overall, and without major uniformly, across the relevant items. It seems likely that there is a gender differences, 76% thought natural foods are healthier, 34% genuine regional difference here. thought they taste better, 19% agreed that they “would rather date someone who eats lots of fruits and veggies rather than lots of meats,” and 4% agreed that “people who eat lots of candy aren’t as Body Mass Index (BMI) nice as people who eat lots of fruits and vegetables.” The standard measure of “heaviness,” BMI (weight in kilograms School and religion are the only variables that show significant divided by height in meters squared) is known to vary by gender, effects (Table 4), and these are very modest compared with the and not surprisingly, it does in this sample (see Tables 3 and 4). gender effects described for the other factors. University of Texas The lowest mean BMI was in University of Pennsylvania women men, very surprisingly, are highest on the natural/vegetarian factor, (21.3) and the highest in Pennsylvania State men (24.3). There is as are Hindus (and to a lesser extent, South Asians). It is no a modest school effect, F(5, 2113) ⫽ 3.73, p ⬍ .01, with the accident that Hindus are highly represented in the University of University of Pennsylvania lowest and University of Wisconsin Texas sample. University of Texas men scored higher than men at highest. There is a more substantial religion effect, F(6, other schools across the board on the items loading on 2056) ⫽ 4.03, p ⬍ .001, with Hindus lowest, a yet more substantial natural/vegetarian. race effect, F(6, 2081) ⫽ 12.35, p ⬍ .001, with South Asians Overall negative/ambivalent reaction to food: Combined factor lowest and African Americans highest. Surprisingly, given the score. We created a mean of the six factor scores for each indications in the literature that fatness decreases as one moves to participant, as a general measure of negative–positive attitudes to higher classes, the correlation between self-reported social class food (Tables 3 and 4). For each of the component factor scores, a and BMI was only ⫺.09 ( p ⬍ .01), for women and ⫺.05 (ns) for positive value indicates food concern and worries and a greater men. food– health and natural orientation, and a negative score indicates a more food–weight positive attitude, with a greater focus on the pleasure of eating. As would be expected from the component Discussion results, by far the biggest effect is for gender, F(1, 2160) ⫽ 328.43, p ⬍ .001, but with much smaller but significant effects for race, Our findings confirm and refine many prior indications that religion and school (Table 4). The gender effect is large (.17 z units American women are very concerned about diet and their body for women, ⫺.25 for men, for a total difference of .42 z units). For appearance (e.g. Rodin, 1992; Rodin et al., 1985; Rozin & Fallon, school, the maximum difference is between the University of 1988; Schwartz, 1986). Our findings also extend this, in keeping Texas (⫺.06) and Pennsylvania State University (.07), but the with a much smaller literature, to great concern about diet and latter results in large part from the fact that the proportion of men health over and above issues of calories and weight (Rozin et al., is lowest at Pennsylvania State. For religion (not confounded by 1999). As has been noted (Rozin, 1998; Rozin et al., 1999), the gender), the maximum spread is ⫺.07 for Buddhists and .06 for greater concern about health does not follow directly from any- Jews. Men from all six schools had negative factor scores, and thing we know about women’s weight concerns. The longer life women from five had positive scores, with the University of Texas expectancy and lower rate of cardiovascular disease in women as at zero. The extreme scores were .27 for University of Wisconsin compared with men in developed countries indicate that a rela- women and ⫺.28 for Arizona State University men (Table 3). tively greater concern about health in men would be warranted. As Notably, the California women were not the most extreme female it happens, there is direct overlap between reducing calories and group. eating a healthy diet, given the link between body weight and mortality. Furthermore, the most calorically dense foods, fats, are Anomalous Results in the University of Texas Sample also most implicated in heart disease independent of their caloric contribution. However, at least three of the items loading on the There were only two substantial violations of a general gender diet– health concern factor (NUTTASTE, LONGTERM, and pattern in the results. One concerns the very low (male range) BREAD; see Table 2) do not bear directly on weight control. In factor score for University of Texas women on health orientation. addition to the confound between healthy and weight-reducing The other is the very high score for Texas men on the natural/ foods, there may be a link between the two motivations in the vegetarian factor. Because the University of Texas sample is most minds of some women. Adopting a healthy diet may be more different from the others in racial composition and social class, at acceptable in some circles than weight reduction, especially when
140 ROZIN, BAUER, AND CATANESE a woman’s weight does not seem to require dieting. Thus, a BMI for the California students, both the men and women, was the woman might disguise a weight-loss diet as a health diet. second lowest of the six schools, with the University of Pennsyl- Our findings help to define the nature of concern about food and vania lowest for both genders. Because the University of Pennsyl- diet among American college students, indicating that gender vania is a private school, it differs from the others in being differences are very substantial on all factors except the food somewhat more upper middle and upper class and in drawing negativity (importance of food in a positive context and the students from a wider range of locations. Among the five state positive–negative framing of food). It seems that the main male– schools, in accordance with the data we have cited, the California female difference has to do not with the positive potential of food sample is somewhat thinner. as a source of pleasure (given very similar scores on the food Other than the BMI results, our data show no trace of a greater importance positive–negative factor) but the degree of negative weight or health– diet concern among California undergraduates. feelings that conflict with the positive feelings. The male–female The University of California at Santa Barbara would seem to be a difference extends across behaviors, attitudes, and modes of reasonable choice for coastal California populations. Our pilot data thinking. showing a difference on a much smaller sample may reflect a The results provide, after the fact, a justification for the common substantial difference between California undergraduates in gen- practice of generalizing from a student sample at one university to eral and those Californians who come East for their education to a all American students. Previous data, which compared a range of selective private university. It is possible that the West Coast American adults with American college students (Rozin et al., diet– health stereotype is incorrect, perhaps promoted by a very 1999; Rozin & Fallon, 1988), have provided some justification for small but very visible segment of the West Coast population that a generalization from a particular student sample to Americans is devoted to exercise, diet, and health foods. more generally. However, there are severe limits to generalization In contrast to the minimal regional differences within the United from American college students to people in general (Rozin et al., States, the contrast of the United States with France is sharp and 1999). Furthermore, though regional differences in the United highly significant for all of the types of measures except the effect States are rather small, among college students, regional general- of diet on health. The French are less likely to diet or make food ization can only be countenanced with some equalization of or choices dominated by concerns about health; find the pleasure of compensation for racial and socioeconomic differences. eating relatively uncontaminated by worries and guilt; and tend to In general, the factor structure that we extracted from our think of food as a sensory and social experience rather than as a questionnaire is very similar to the only other parallel structure, source of nutrients, toxins, and calories (Rozin et al., 1999). The from the Rozin et al. (1999) study of food in life in four countries. French–American difference appears for both men and women; as The first four factors are about the same, but the eating disorder with Americans, it is the French women who are relatively more and natural/vegetarian factors are not distinct in the four-country concerned about weight, diet, and health than are the men (Rozin sample. Rather, there is a factor suggesting culinary versus et al, 1999). We do not fully understand the origin of the French– nutrition– health thinking about food. American differences in relationships to food. Possible accounts The present study was motivated, in part, by our impression and include the valuation of moderation in France versus excess in the some pilot data suggesting much greater concern about diet– health United States and religious differences, with Catholicism promi- issues in Californians than in people in other geographical areas of nent in France and Protestantism in the United States (Rozin, 1998; the United States. There is some modest evidence for healthier Rozin et al., 1999; Stearns, 1997). Whatever the origins of the habits among Californians. Female California high school students difference, the environment established by French culture seems to have a relative mean weight 96.8% of the entire national standard encourage less food consumption (mediated at least in part by (Mellin, Irwin, & Scully, 1992). This slight support for a Califor- smaller portion size; Rozin et al., in press), a more varied diet nia stereotype is weakly confirmed by data we obtained from the (Drewnowski et al., 1996), perhaps more time spent eating (Rozin marketing department at McDonald’s (personal communication). et al., in press), and perhaps more exercise integrated with the daily We determined the number of McDonald’s outlets in each of the routine. five states covered by our study and divided this into the popula- The findings we report here have implications for the under- tion for each state, yielding a people per McDonald’s measure. We standing of eating disorders. The small minority of the female proposed that a greater health orientation in California might lead college population that is diagnosed with eating disorders might be to a higher ratio in California because McDonald’s might be less considered the top of a large pyramid of degrees of concern about popular there. (Note that the issue here is not whether eating at weight and diet. The normative female concern seems to involve McDonald’s is good for health, but rather that people perceive that issues of both intake and food choice and to extend across behav- fast food is not as healthy as other forms of food.) The ratios we iors, attitudes, and modes of thought. obtained were as follows, in order of increasing people per Mc- We believe it is profitable, from the point of view of under- Donald’s: Wisconsin, 23,290; Texas, 23,405; Arizona, 23,433; standing the obsessions with food of many American women, to Pennsylvania, 24,994; and California, 27,600. The density is low- extend our understanding of relationships to food beyond the est, by a modest margin, in California. A survey on food intake narrow domains of dieting and weight concern. We have made a covering various regions of the United States in the 1989 –1991 beginning in creating the broader canvas in our prior cross-cultural period showed that people from the West Coast consumed 4.8 work (Rozin et al., 1999) and in the current study. Our emphasis servings of meat per day compared with 5.3 in the Northeast on default thinking (see also analysis of free associations to food in and 5.2 in both the South and Midwest, as well as increased Rozin, Kurzer, & Cohen, 2002) introduces a dimension of poten- consumption of fruits and vegetables on the West Coast (Cleve- tial interest. However, there is much work to be done in expanding land, Cook, Kreb-Smith, & Friday, 1997). In the present study, the our vision of the many ways in which food functions and in
FOOD AND LIFE, PLEASURE AND WORRY 141 refining the measures we have developed. Also, we have much to Rozin, P. (1998). Food is fundamental, fun, frightening, and far-reaching. learn about the development of the American female pattern of Social Research, 66, 9 –30. relations to food. Most of the male–female difference seems to Rozin, P., Ashmore, M. B., & Markwith, M. (1996). Lay American emerge at puberty, so this would seem to be a natural transition for conceptions of nutrition: Dose insensitivity, categorical thinking, conta- gion, and the monotonic mind. Health Psychology, 15, 438 – 447. scholarly attention. Rozin, P., & Fallon, A. E. (1988). Body image, attitudes to weight, and misperceptions of figure preferences of the opposite sex: A comparison References of males and females in two generations. Journal of Abnormal Psychol- ogy, 97, 342–345. Bauer, R., & Catanese, D. (1996). Concerns about weight and body image Rozin, P., Fischler, C., Imada, S., Sarubin, A., & Wrzesniewski, A. (1999). in Californian and Eastern freshman females at the University of Penn- Attitudes to food and the role of food in life in the U.S.A., Japan, sylvania. Unpublished manuscript, University of Pennsylvania. Flemish Belgium and France: Possible implications for the diet– health Cleveland, L. E., Cook, D. A., Kreb-Smith, S. M., & Friday, J. (1997). debate. Appetite, 33, 163–180. Method for assessing food intakes in terms of servings based on food Rozin, P., Kabnick, K., Pete, E., Fischler, C., & Shields, C. (in press). The guidance. American Journal of Clinical Nutrition, 65(Suppl. 4) , 1245S– ecology of eating: Part of the French paradox results from lower food 1263S. intake in French than Americans, because of smaller portion sizes. Drewnowski, A., Henderson, S. A., Shore, A. B., Fischler, C., Preziosi, P., Psychological Science. & Hercberg, S. (1996). Diet quality and dietary diversity in France: Rozin, P., Kurzer, N., & Cohen, A. (2002). Free associations to “food”: Implications for the French paradox. Journal of the American Dietetic The effects of gender, generation, and culture. Journal of Research in Association, 96, 663– 669. Personality, 36, 419 – 441. Fischler, C. (1990). L’homnivore [The homnivore]. Paris: Odile. Schwartz, H. (1986). Never satisfied: A cultural history of diets, fantasies, Kass, L. (1994). The hungry soul. New York: Free Press. and fat. New York: Free Press. Mellin, L. M., Irwin, C. E., Jr., & Scully, S. (1992). Prevalence of SPSS. (1999). Systat9 [Computer software]. Chicago: Author. disordered eating in girls: A survey of middle-class children. Journal of Stearns, P. N. (1997). Fat history: Bodies and beauty in the modern West. the American Dietetic Association, 92, 851– 853. New York: New York University Press. Polivy, J., & Herman, C. P. (2002). If at first you don’t succeed: False Wadden, T. A., Brownell, K. D., & Foster, G. D. (2002). Obesity: Re- hopes of self-change. American Psychologist, 57, 677– 689. sponding to the global epidemic. Journal of Consulting and Clinical Renaud, S., & de Lorgeril, M. (1989). Dietary lipids and their relation to Psychology, 70, 510 –525. ischaemic heart disease: From epidemiology to prevention. Journal of Wadden, T. A., & Foster, G. D. (2000). Behavioral treatment of obesity. Internal Medicine, 225(Suppl. 1), 39 – 46. Medical Clinics of North America, 84, 441– 460. Rodin, J. (1992). Body traps. New York: Morrow. Rodin, J., Silberstein, L. R., & Striegel-Moore, R. (1985). Women and weight: A normative discontent. In T. B. Sonderegger (Ed.), Nebraska Received June 19, 2002 Symposium on Motivation: Vol. 32. Psychology and gender (pp. 267– Revision received November 4, 2002 307). Lincoln: University of Nebraska Press. Accepted November 6, 2002 䡲
You can also read
NEXT SLIDES ... Cancel